DEPOOL Benchmark Evaluates Temporal Aggregation for Depression Detection
July 2, 2026
DEPOOL introduces a controlled benchmark comparing six aggregation architectures against six speech backbones for depression detection. The study reveals that a third of the 72 configurations fail by collapsing into single-class predictions, highlighting backbone sensitivity.
HOW THIS AFFECTS YOU
●
researcherYou can use this benchmark to isolate the effects of aggregation methods from the choice of encoder.
●
healthThis research identifies critical failure modes in automated speech-based mental health diagnostics.